Obsidian General Meeting Processor

Process regular meeting notes (non-standup meetings) with attendee extraction, transcript cleanup, and summary generation.

Capabilities

1. •Attendee Extraction - Process Teams screenshots to identify participants
2. •People Profile Creation - Generate linked profiles with avatars and aliases
3. •Known People Lookup - Find attendees mentioned in content when no screenshot available
4. •Transcript Cleaning - Handle 4 different Teams transcript formats
5. •Copilot Summary Detection - Recognize and preserve Microsoft Copilot summaries
6. •Summary Generation - Create structured meeting notes from transcripts
7. •Summary Augmentation - Enhance existing summaries with transcript details

Processing Workflow

Step 1: Extract Attendees

Option A: From Screenshot (Preferred)

If meeting has attendee screenshot reference (![[SCR-*.png]]):

1. •View image from ~/Documents/Obsidian/HPE/Media/
2. •Extract names in "Last, First" format
3. •Extract individual avatars (40x40 pixels) using sips:
   bash

   cd ~/Documents/Obsidian/HPE/Media
   # First attendee
   sips -c 40 40 --cropOffset 12 12 SCR-YYYYMMDD-xxxxx.png --out Last-First-avatar.png
   # Second attendee (~54 pixels down)
   sips -c 40 40 --cropOffset 66 12 SCR-YYYYMMDD-xxxxx.png --out Last-First-avatar.png
   

4. •Create People profiles with avatars
5. •Update ## Attendees section with linked list
6. •Remove screenshot reference after extraction

Option B: Known People Lookup (No Screenshot)

If no screenshot available:

1. •Extract potential names from meeting content:
   bash

   grep -oE "\b[A-Z][a-z]+\b" "meeting-file.md" | sort -u
   

2. •Search People profiles for matches:
   bash

   cd ~/Documents/Obsidian/HPE/People
   for name in Yogesh Kyu Darra; do
     ls -1 *.md | grep -i "$name"
   done
   

3. •Verify matches by checking aliases
4. •Build attendee list from verified matches

Step 2: Create People Profiles

Check ~/Documents/Obsidian/HPE/People/{Last}, {First}.md - create if missing:

---
aliases:
  - {First} {Last}
tags:
  - People
---
![[{Last}-{First}-avatar.png]]

Important:

• •Only add avatar if profile doesn't already have one
• •Check for existing ![[ image reference first
• •Use exact format: filename {Last}, {First}.md, alias {First} {Last}

Step 3: Update Meeting Attendees

Replace image reference with linked attendee list:

## Attendees

## In Meeting (5)
- [[Tronkowski, Kevin|Kevin Tronkowski]]
- [[Pahwa, Kashish|Kashish Pahwa]]

## Invited/Other Participants (3)
- [[Luna, Gabriella|Gabriella Luna]]

Include counts in section headers: ## In Meeting (5)

Step 4: Check for Copilot Summary

Check if ## Copilot Summary section exists with content:

• •If present: Skip transcript cleanup and summary generation
• •If absent or empty: Proceed with transcript processing

Copilot Summary Characteristics:

• •Section header: ## Copilot Summary
• •Structured subsections: # Decisions, # Open question, # Agenda, # Meeting notes
• •Already formatted (no cleanup needed)
• •Often present WITHOUT transcript

When Copilot Summary exists:

1. •✅ Process attendees normally
2. •✅ Keep Copilot-generated content intact
3. •❌ Skip transcript cleanup
4. •❌ Skip summary generation
5. •✅ Optionally augment if transcript also present

Step 5: Clean Transcript (if no Copilot Summary)

Use appropriate script based on transcript format:

Format 1: Direct Teams Paste

Characteristics:

• •Speaker names on own line: Last, First
• •Verbose timestamp: 0 minutes 3 seconds0:03
• •Profile image URLs
• •Separator lines ____

Script:

python3 ~/.copilot/skills/obsidian-meeting-general/scripts/clean_transcript.py "meeting-file.md"

Format 2: Downloaded from Teams

Characteristics:

• •Combined speaker/timestamp: **Bennett, Ryan** 0:03
• •Embedded image references
• •Meeting header (title, date, duration)

Script:

python3 ~/.copilot/skills/obsidian-meeting-general/scripts/clean_transcript_downloaded.py "meeting-file.md"

Format 3: .docx Exported

Characteristics:

• •Plain text (no bold): Last, First timestamp content
• •Leading space before name
• •No image references

Script:

python3 ~/.copilot/skills/obsidian-meeting-general/scripts/clean_transcript_docx.py "meeting-file.md"

Format 4: Simple/Generic

Characteristics:

• •Basic formatting, minimal metadata

Script:

python3 ~/.copilot/skills/obsidian-meeting-general/scripts/clean_transcript_simple.py "meeting-file.md"

Step 6: Generate Summary (if no Copilot Summary)

Analyze cleaned transcript and create structured meeting notes:

If existing summary sections found (e.g., # Decisions, # Open questions):

• •Augment with additional detail from transcript
• •Preserve existing structure

If no existing summary:

• •Insert new structured summary in ## Notes section
• •Include: Decisions, Action Items, Open Questions, Key Discussion Points

Finding Meeting Files

# Today's meetings
find ~/Documents/Obsidian/HPE/Meetings -name "$(date +%Y-%m-%d)*.md"

# Specific date
find ~/Documents/Obsidian/HPE/Meetings -name "2026-01-05*.md"

# By name (case-insensitive)
find ~/Documents/Obsidian/HPE/Meetings -iname "*aruba*interlock*.md"

# Date range (last week)
find ~/Documents/Obsidian/HPE/Meetings -name "*.md" -newermt "$(date -v-7d +%Y-%m-%d)" ! -newermt "$(date +%Y-%m-%d)"

Common Usage Patterns

"Process the meeting with transcript"

• •Extract attendees (screenshot or known people lookup)
• •Create missing People profiles
• •Clean transcript (detect format automatically)
• •Generate summary from cleaned transcript

"Process today's aruba meeting"

• •Find matching meeting file
• •Full workflow: attendees → profiles → transcript → summary

"Just clean up the transcript"

• •Skip attendee processing
• •Only clean transcript section
• •Detect format and use appropriate script

"Find the people from obsidian people and list them"

• •When no screenshot available
• •Extract first names from content
• •Search People profiles
• •Update attendee list with found profiles

Validation Checklist

After processing, verify:

• • ## Attendees updated with participants (actual or from content)
• • People profiles created for new attendees (with avatars if screenshot)
• • ## Transcript cleaned (if present and no Copilot Summary)
• • ## Notes or ## Copilot Summary contains meeting summary
• • All links use Obsidian format: [[Last, First|First Last]]
• • Screenshot reference removed after extraction
• • Existing Copilot Summaries preserved (not overwritten)

Important Rules

• •Always check existing profiles before creating
• •Preserve existing profile content when found
• •Use exact format: filename {Last}, {First}.md, alias {First} {Last}
• •Include counts in section headers: ## In Meeting (5)
• •Check for Copilot Summary before cleaning transcript
• •Preserve Copilot-generated summaries
• •Only generate summaries when no Copilot Summary exists
• •Remove screenshot reference after attendee extraction

Directory Structure

~/Documents/Obsidian/HPE/
├── Meetings/YYYY-MM-DD - name.md
├── People/{Last}, {First}.md
└── Media/SCR-YYYYMMDD-xxxxx.png

Additional Resources

• •scripts/README.md - Documentation for transcript cleaning scripts
• •examples.md - Complete meeting examples
• •reference.md - Obsidian markdown syntax reference